The period branch of estimateArchiveSize ran a single unpaginated
document read with one flat IN() over every posted entry id in the
year: past the PostgREST row cap it silently undercounts, and past a
few hundred entry ids the URL itself blows up. Chunk the id filter
(CHILD_FK_CHUNK) and paginate every read with fetchAllRows, mirroring
what writeDocuments already did (the ZIP content was never affected).
The dialog now says per scope which documents are counted: full
history includes unlinked inbox/receipt documents, a single year only
those linked to posted vouchers. Without that line, a company with
many unlinked receipts reads the count gap as a pagination bug.
Co-authored-by: Jakob Wennberg <311770904+jakobwennberg-oss@users.noreply.github.com>
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Jakob Wennberg
2eb3441244